9:00 AM - 10:30 AM
[ACG37-P05] Improvement of GSMaP_MWR by CDF matching with respect to GPM/DPR observations
Keywords:microwave radiometer, radar, precipitation, cumulative distribution function matching
This study aims to improve the accuracy of the microwave radiometer observations in the Global Satellite Mapping of Precipitation (GSMaP_MWR) by employing cumulative distribution matching (CDF) matching with respect to precipitation data measured by the Global Precipitation Measurement Mission core satellite’s Dual-frequency Precipitation Radar (GPM/DPR). First, CDFs of hourly precipitation for each 0.1-degrees grid and the surrounding ±5 degrees domain were produced for both GSMaP_MWR and DPR. Here, we used the data samples from 2015 to 2020, and produced the CDFs differently for each season (DJF, MAM, JJA, SON). Then, the data samples of each grid in the GSMaP_MWR were corrected by CDF matching referring to the CDFs of DPR. Finally, we verified the original and corrected data of GSMaP_MWR against Radar-AMeDAS precipitation over Japan.
Our results show that the CDF matching decreases differences between GSMaP_MWR and DPR for all seasons, especially on land in the Northern hemisphere, in the Southern hemisphere, and throughout the globe for MAM and JJA, DJF, and SON, respectively. These results indicate that the impacts of CDF matching varies seasonally for each region.
Furthermore, our results show that for DJF, MAM, and SON, CDF matching successfully improved the accuracy of the GSMaP_MWR in Japan. The monthly mean absolute error (MAE) between the GSMaP_MWR and Radar-AMeDAS decreased after the CDF matching in months except for JJA. In addition, the difference in the seasonal average of precipitation amount between the GSMaP_MWR and Radar-AMeDAS became smaller for these periods. On the other hand, the monthly MAE increased and the difference in the seasonal average of precipitation amount increased after the CDF matching in JJA. This is presumed to have occurred owing to the large difference between the data samples in DPR and in Radar-AMeDAS for this season. However, the difference in the average annual precipitation amount between the GSMaP_MWR and Radar-AMeDAS decreased, indicating that the CDF matching method in this study improved the accuracy of the GSMaP_MWR overall.
Our results show that the CDF matching decreases differences between GSMaP_MWR and DPR for all seasons, especially on land in the Northern hemisphere, in the Southern hemisphere, and throughout the globe for MAM and JJA, DJF, and SON, respectively. These results indicate that the impacts of CDF matching varies seasonally for each region.
Furthermore, our results show that for DJF, MAM, and SON, CDF matching successfully improved the accuracy of the GSMaP_MWR in Japan. The monthly mean absolute error (MAE) between the GSMaP_MWR and Radar-AMeDAS decreased after the CDF matching in months except for JJA. In addition, the difference in the seasonal average of precipitation amount between the GSMaP_MWR and Radar-AMeDAS became smaller for these periods. On the other hand, the monthly MAE increased and the difference in the seasonal average of precipitation amount increased after the CDF matching in JJA. This is presumed to have occurred owing to the large difference between the data samples in DPR and in Radar-AMeDAS for this season. However, the difference in the average annual precipitation amount between the GSMaP_MWR and Radar-AMeDAS decreased, indicating that the CDF matching method in this study improved the accuracy of the GSMaP_MWR overall.